Cajal-Retzius cells Trajectory

Load libraries

library(Seurat)
library(princurve)
library(monocle)
library(gprofiler2)
library(seriation)
library(Matrix)
library(dplyr)
library(RColorBrewer)
library(ggplot2)
library(ggExtra)
library(cowplot)
library(wesanderson)

#Set ggplot theme as classic
theme_set(theme_classic())

Load the full dataset

Hem.data <- readRDS("../QC.filtered.cells.RDS")
Idents(Hem.data) <- Hem.data$Cell_ident

DimPlot(object = Hem.data,
        group.by = "Cell_ident",
        reduction = "spring",
        cols = c("#83c3b8", #"ChP"
                 "#009fda", #"ChP_progenitors"
                 "#68b041", #"Dorso-Medial_pallium"
                 "#e46b6b", #"Hem"
                 "#e3c148", #"Medial_pallium"
                 "#b7d174", #2
                 "grey40", #4
                 "black", #5
                 "#3e69ac" #"Thalamic_eminence"
                 ))

Differentiating neurons trajectory

Neurons.data <-  subset(Hem.data, idents = c("seurat_clusters_2"))

DimPlot(Neurons.data ,
        reduction = "spring",
        pt.size = 1,
        cols =  c("#b7d174")) + NoAxes()

Split Pallial from Cajal-Retzius cells

p1 <- FeaturePlot(object = Neurons.data ,
            features = c("BP_signature1","LN_signature1"),
            pt.size = 0.5,
            cols = rev(brewer.pal(10,"Spectral")),
            reduction = "spring",
            order = T) & NoAxes()

p2 <- FeaturePlot(object = Neurons.data ,
            features = c("Foxg1", "Trp73"),
            pt.size = 0.5,
            cols = c("grey90", brewer.pal(9,"YlGnBu")),
            reduction = "spring",
            order = T) & NoAxes()

p1 / p2

Separation between the 2 lineage seems straightforward. We use louvain clustering to split the two.

Neurons.data <- RunPCA(Neurons.data, verbose = FALSE)

Neurons.data <- FindNeighbors(Neurons.data,
                              dims = 1:10,
                              k.param = 8)

Neurons.data <- FindClusters(Neurons.data, resolution = 0.05)

## Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
## 
## Number of nodes: 2835
## Number of edges: 56608
## 
## Running Louvain algorithm...
## Maximum modularity in 10 random starts: 0.9640
## Number of communities: 2
## Elapsed time: 0 seconds

DimPlot(Neurons.data,
        reduction = "spring",
        cols = c("#cc391b","#026c9a"),
        pt.size = 0.5) & NoAxes()

Neurons.data$Lineage <- sapply(as.numeric(Neurons.data$SCT_snn_res.0.05),
                               FUN = function(x) {x= c("Hem","Pallial")[x]})

DimPlot(object = Neurons.data,
        group.by = "Lineage",
        reduction = "spring",
        cols = c("#cc391b","#026c9a"),
        pt.size = 0.5) & NoAxes()

Fit principale curve on the two lineages

Cajal-Retzius cells

Trajectories.Hem <- Neurons.data@meta.data %>%
                    select("Barcodes", "nUMI", "Spring_1", "Spring_2", "Lineage") %>%
                    filter(Lineage == "Hem")

fit <- principal_curve(as.matrix(Trajectories.Hem[,c("Spring_1", "Spring_2")]),
                       smoother='lowess',
                       trace=TRUE,
                       f = .7,
                       stretch=0)

## Starting curve---distance^2: 45804778678
## Iteration 1---distance^2: 27732113
## Iteration 2---distance^2: 27728318

#The principal curve smoothed
Hem.pc.line <- as.data.frame(fit$s[order(fit$lambda),]) 

#Pseudotime score
Trajectories.Hem$PseudotimeScore <- fit$lambda/max(fit$lambda)

if (cor(Trajectories.Hem$PseudotimeScore, Neurons.data@assays$SCT@data['Hmga2', Trajectories.Hem$Barcodes]) > 0) {
  Trajectories.Hem$PseudotimeScore <- -(Trajectories.Hem$PseudotimeScore - max(Trajectories.Hem$PseudotimeScore))
}

Pallial neurons

Trajectories.Pallial <- Neurons.data@meta.data %>%
                        select("Barcodes", "nUMI", "Spring_1", "Spring_2", "Lineage") %>%
                        filter(Lineage == "Pallial")

fit <- principal_curve(as.matrix(Trajectories.Pallial[,c("Spring_1", "Spring_2")]),
                       smoother='lowess',
                       trace=TRUE,
                       f = .7,
                       stretch=0)

## Starting curve---distance^2: 26984853690
## Iteration 1---distance^2: 22153700
## Iteration 2---distance^2: 22179462
## Iteration 3---distance^2: 22180297

#The principal curve smoothed
Pallial.pc.line <- as.data.frame(fit$s[order(fit$lambda),])

#Pseudotime score
Trajectories.Pallial$PseudotimeScore <- fit$lambda/max(fit$lambda)

if (cor(Trajectories.Pallial$PseudotimeScore, Neurons.data@assays$SCT@data['Hmga2', Trajectories.Pallial$Barcodes]) > 0) {
  Trajectories.Pallial$PseudotimeScore <- -(Trajectories.Pallial$PseudotimeScore - max(Trajectories.Pallial$PseudotimeScore))
}

Combine the two trajectories’ data

Trajectories.neurons <- rbind(Trajectories.Pallial, Trajectories.Hem)

cols <- brewer.pal(n =11, name = "Spectral")

ggplot(Trajectories.neurons, aes(Spring_1, Spring_2)) +
  geom_point(aes(color=PseudotimeScore), size=2, shape=16) + 
  scale_color_gradientn(colours=rev(cols), name='Speudotime score') +
  geom_line(data=Pallial.pc.line, color="#026c9a", size=0.77) +
  geom_line(data=Hem.pc.line, color="#cc391b", size=0.77)

Plot pan-neuronal genes along this axis

Neurons.data <- NormalizeData(Neurons.data, normalization.method = "LogNormalize", scale.factor = 10000, assay = "RNA")

# Neurog2
p1 <- FeaturePlot(object = Neurons.data,
            features = c("Neurog2"),
            pt.size = 0.5,
            cols = c("grey90", brewer.pal(9,"YlGnBu")),
            reduction = "spring",
            order = T) & NoAxes()

Trajectories.neurons$Neurog2 <- Neurons.data@assays$RNA@data["Neurog2", Trajectories.neurons$Barcodes]

p2 <- ggplot(Trajectories.neurons, aes(x= PseudotimeScore, y= Neurog2)) +
        geom_point(aes(color= Lineage), size=0.5) +
        scale_color_manual(values= c("#cc391b", "#026c9a")) +
        geom_smooth(method="loess", n= 50, aes(color= Lineage)) +
        ylim(0,NA)

# Tbr1 
p3 <- FeaturePlot(object = Neurons.data ,
            features = c("Tbr1"),
            pt.size = 0.5,
            cols = c("grey90", brewer.pal(9,"YlGnBu")),
            reduction = "spring",
            order = T) & NoAxes()
Trajectories.neurons$Tbr1 <- Neurons.data@assays$RNA@data["Tbr1", Trajectories.neurons$Barcodes]

p4 <- ggplot(Trajectories.neurons, aes(x= PseudotimeScore, y= Tbr1)) +
        geom_point(aes(color= Lineage), size=0.5) +
        scale_color_manual(values= c("#cc391b", "#026c9a")) +
        geom_smooth(method="loess", n= 50, aes(color= Lineage)) +
        ylim(0,NA)

# Mapt 
p5 <- FeaturePlot(object = Neurons.data ,
            features = c("Mapt"),
            pt.size = 0.5,
            cols = c("grey90", brewer.pal(9,"YlGnBu")),
            reduction = "spring",
            order = T) & NoAxes()

Trajectories.neurons$Mapt <- Neurons.data@assays$RNA@data["Mapt", Trajectories.neurons$Barcodes]

p6 <- ggplot(Trajectories.neurons, aes(x= PseudotimeScore, y= Mapt)) +
        geom_point(aes(color= Lineage), size=0.5) +
        scale_color_manual(values= c("#cc391b", "#026c9a")) +
        geom_smooth(method="loess", n= 50, aes(color= Lineage)) +
        ylim(0,NA)

p1 + p2 + p3 + p4 + p5 + p6 + patchwork::plot_layout(ncol = 2)

Shift Pseudotime in both lineage

Since we observe the first 25% of both trajectories are occupied by few, likely progenitor cells, we shift this cell along the axis

Pseudotime.intervals <- Trajectories.neurons%>%
                          select(Lineage, PseudotimeScore) %>%
                          mutate(Pseudotime.bins = cut(Trajectories.neurons$PseudotimeScore, seq(0, max(Trajectories.neurons$PseudotimeScore) + 0.05, 0.05), dig.lab = 2, right = FALSE)) %>%
                          group_by(Lineage, Pseudotime.bins) %>%
                          summarise(n=n())

ggplot(Pseudotime.intervals, aes(x=Pseudotime.bins, y=n, fill=Lineage)) +
        geom_bar(stat = "identity", width = 0.90) +
        theme(axis.text.x = element_text(angle = 45, hjust=1))+
        scale_fill_manual(values= c("#cc391b", "#026c9a"))

score <- sapply(Trajectories.neurons$PseudotimeScore,
                FUN = function(x) if (x <= 0.2) {x= 0.2} else { x=x })

Trajectories.neurons$PseudotimeScore.shifted <- (score - min(score)) / (max(score) - min(score))

# Neurog2
p1 <- FeaturePlot(object = Neurons.data ,
            features = c("Neurog2"),
            pt.size = 0.5,
            cols = c("grey90", brewer.pal(9,"YlGnBu")),
            reduction = "spring",
            order = T) & NoAxes()

p2 <- ggplot(Trajectories.neurons, aes(x= PseudotimeScore.shifted, y= Neurog2)) +
        geom_point(aes(color= Lineage), size=0.5) +
        scale_color_manual(values= c("#cc391b", "#026c9a")) +
        geom_smooth(method="loess", n= 50, aes(color= Lineage)) +
        ylim(0,NA)

# Tbr1 
p3 <- FeaturePlot(object = Neurons.data ,
            features = c("Tbr1"),
            pt.size = 0.5,
            cols = c("grey90", brewer.pal(9,"YlGnBu")),
            reduction = "spring",
            order = T) & NoAxes()

p4 <- ggplot(Trajectories.neurons, aes(x= PseudotimeScore.shifted, y= Tbr1)) +
        geom_point(aes(color= Lineage), size=0.5) +
        scale_color_manual(values= c("#cc391b", "#026c9a")) +
        geom_smooth(method="loess", n= 50, aes(color= Lineage)) +
        ylim(0,NA)

# Mapt 
p5 <- FeaturePlot(object = Neurons.data ,
            features = c("Mapt"),
            pt.size = 0.5,
            cols = c("grey90", brewer.pal(9,"YlGnBu")),
            reduction = "spring",
            order = T) & NoAxes()

p6 <- ggplot(Trajectories.neurons, aes(x= PseudotimeScore.shifted, y= Mapt)) +
        geom_point(aes(color= Lineage), size=0.5) +
        scale_color_manual(values= c("#cc391b", "#026c9a")) +
        geom_smooth(method="loess", n= 50, aes(color= Lineage)) +
        ylim(0,NA)

p1 + p2 + p3 + p4 + p5 + p6 + patchwork::plot_layout(ncol = 2)

ggplot(Trajectories.neurons, aes(x= PseudotimeScore.shifted, y= nUMI/10000)) +
        geom_point(aes(color= Lineage), size=0.5) +
        scale_color_manual(values= c("#cc391b", "#026c9a")) +
        geom_smooth(method="loess", n= 50, aes(color= Lineage)) +
        ylim(0,NA)

rm(list = ls()[!ls() %in% c("Trajectories.neurons")])

Load progenitors with cell cycle trajectory fitted

Progenitors.data <- readRDS("../ProgenitorsDiversity/Progenitors.RDS")

table(Progenitors.data$Cell_ident)

## 
## Dorso-Medial_pallium                  Hem       Medial_pallium 
##                 3451                 1954                 2719

To balance the number of progenitors in both domain we will only work with Hem and Medial_pallium annotated cells. Since we are using pallial cell to contrast CR specific trajectory we think this approximation will not significantly affect our analysis.

Progenitors.data <-  subset(Progenitors.data, idents = c("Hem", "Medial_pallium"))

p1 <- DimPlot(Progenitors.data,
        reduction = "spring",
        pt.size = 0.5,
        cols =  c("#e3c148", "#e46b6b")) + NoAxes()

p2 <- FeaturePlot(object = Progenitors.data,
            features = "Angle.cc",
            pt.size = 0.5,
            cols = rev(colorRampPalette(brewer.pal(n =10, name = "Spectral"))(100)),
            reduction = "spring",
            order = T) & NoAxes()

p3 <- DimPlot(object = Progenitors.data,
        group.by = "Revelio.phase",
        pt.size = 0.5,
        reduction = "spring",
        cols =  c(wes_palette("FantasticFox1")[1:3],"grey40",wes_palette("FantasticFox1")[5])) & NoAxes()

p1 + p2 + p3  + patchwork::plot_layout(ncol = 2)

Combined progenitors and neurons along Pseudotime

# Start with neurons data
Trajectories.all <- Trajectories.neurons %>% select(Barcodes, nUMI, Spring_1, Spring_2, Lineage)

Trajectories.all$Pseudotime <- Trajectories.neurons$PseudotimeScore.shifted + 1
Trajectories.all$Phase <- NA

# Add progenitors data
Trajectories.progenitors <- Progenitors.data@meta.data %>%
                              select(Barcodes, nUMI, Spring_1, Spring_2) %>% 
                              mutate(Lineage= ifelse(Progenitors.data$Cell_ident == "Medial_pallium", "Pallial", "Hem") ,
                                     Pseudotime= Progenitors.data$Angle.cc,
                                     Phase = Progenitors.data$Revelio.phase)

Trajectories.all <- rbind(Trajectories.all, Trajectories.progenitors)

Trajectories.all$Phase <- factor(Trajectories.all$Phase, levels = c("G1.S", "S", "G2", "G2.M", "M.G1"))

p1 <- ggplot(Trajectories.all, aes(Spring_1, Spring_2)) +
        geom_point(aes(color=Pseudotime), size=0.5) + 
        scale_color_gradientn(colours=rev(brewer.pal(n =11, name = "Spectral")), name='Speudotime score')

p2 <- ggplot(Trajectories.all, aes(Spring_1, Spring_2)) +
        geom_point(aes(color= Lineage), size=0.5) +
        scale_color_manual(values= c("#cc391b", "#026c9a"))

p1 + p2

p1 <- ggplot(Trajectories.all, aes(x= Pseudotime, y= nUMI/10000)) +
        geom_point(aes(color= Phase), size=0.5) +
        scale_color_manual(values= c(wes_palette("FantasticFox1")[1:3],"grey40",wes_palette("FantasticFox1")[5])) +
        geom_smooth(method="loess", n= 50, fill="grey") +
        ylim(0,NA)

p2 <- ggplot(Trajectories.all, aes(x= Pseudotime, y= nUMI/10000)) +
        geom_point(aes(color= Lineage), size=0.5) +
        scale_color_manual(values= c("#cc391b", "#026c9a")) +
        geom_smooth(method="loess", n= 50, fill="grey") +
        ylim(0,NA)

p1 + p2

rm(list = ls()[!ls() %in% c("Trajectories.all")])

Subset the full dataset Seurat object

Hem.data <- readRDS("../QC.filtered.cells.RDS")

Neuro.trajectories <- CreateSeuratObject(counts = Hem.data@assays$RNA@data[, Trajectories.all$Barcodes],
                                         meta.data = Trajectories.all)

spring <- as.matrix(Neuro.trajectories@meta.data %>% select("Spring_1", "Spring_2"))
  
Neuro.trajectories[["spring"]] <- CreateDimReducObject(embeddings = spring, key = "Spring_", assay = DefaultAssay(Neuro.trajectories))

p1 <- FeaturePlot(object = Neuro.trajectories,
            features = "Pseudotime",
            pt.size = 1,
            cols = rev(colorRampPalette(brewer.pal(n =11, name = "Spectral"))(100)),
            reduction = "spring",
            order = T) & NoAxes()

p2 <- DimPlot(object = Neuro.trajectories,
        group.by = "Lineage",
        pt.size = 1,
        reduction = "spring",
        cols =  c("#cc391b", "#026c9a")) & NoAxes()


p3 <- DimPlot(object = Neuro.trajectories,
        group.by = "Phase",
        pt.size = 1,
        reduction = "spring",
        cols =  c(wes_palette("FantasticFox1")[1:3],"grey40",wes_palette("FantasticFox1")[5])) & NoAxes()

p1 + p2 + p3

rm(list = ls()[!ls() %in% c("Neuro.trajectories")])

Normalization

Neuro.trajectories<- NormalizeData(Neuro.trajectories, normalization.method = "LogNormalize", scale.factor = 10000, assay = "RNA")

Neuro.trajectories <- FindVariableFeatures(Neuro.trajectories, selection.method = "disp", nfeatures = 3000, assay = "RNA")

Plot some genes along pseudotime

source("../Functions/functions_GeneTrends.R")

Plot.Genes.trend(Seurat.data= Neuro.trajectories,
                 group.by = "Lineage",
                 genes= c("Gas1","Sox2",
                          "Neurog2", "Btg2",
                          "Tbr1", "Mapt",
                          "Trp73", "Foxg1"))

Plot.Genes.trend(Seurat.data= Neuro.trajectories,
                 group.by = "Lineage",
                 genes= c("Gmnc", "Mcidas",
                          "Foxj1", "Trp73",
                          "Lhx1", "Cdkn1a"))

Plot.Genes.trend(Seurat.data= Neuro.trajectories,
                 group.by = "Lineage",
                 genes= c("Mki67", "Top2a",
                          "H2afx", "Cdkn1c"))

Use monocle2 to model gene expression along cycling axis

Initialize a monocle object

# Transfer metadata
meta.data <- data.frame(Barcode= Neuro.trajectories$Barcodes,
                        Lineage= Neuro.trajectories$Lineage,
                        Pseudotime= Neuro.trajectories$Pseudotime,
                        Phase= Neuro.trajectories$Phase)

Annot.data  <- new('AnnotatedDataFrame', data = meta.data)

# Transfer counts data
var.genes <- Neuro.trajectories[["RNA"]]@var.features
count.data = data.frame(gene_short_name = rownames(Neuro.trajectories[["RNA"]]@data[var.genes,]),
                        row.names = rownames(Neuro.trajectories[["RNA"]]@data[var.genes,]))

feature.data <- new('AnnotatedDataFrame', data = count.data)

# Create the CellDataSet object including variable genes only
gbm_cds <- newCellDataSet(Neuro.trajectories[["RNA"]]@counts[var.genes,],
                          phenoData = Annot.data,
                          featureData = feature.data,
                          lowerDetectionLimit = 0,
                          expressionFamily = negbinomial())

gbm_cds <- estimateSizeFactors(gbm_cds)
gbm_cds <- estimateDispersions(gbm_cds)
gbm_cds <- detectGenes(gbm_cds, min_expr = 0.1)

rm(list = ls()[!ls() %in% c("Neuro.trajectories", "gbm_cds", "Gene.Trend", "Plot.Genes.trend")])
gc()

##             used  (Mb) gc trigger   (Mb)  max used   (Mb)
## Ncells   3588903 191.7    6211198  331.8   6211198  331.8
## Vcells 102208537 779.8  404496316 3086.1 612662844 4674.3

Test each gene trend over pseudotime score

Find genes DE along pseudomaturation axis

pseudo.maturation.diff <- differentialGeneTest(gbm_cds[fData(gbm_cds)$num_cells_expressed > 80,], 
                                                 fullModelFormulaStr = "~sm.ns(Pseudotime, df = 3)*Lineage", 
                                                 reducedModelFormulaStr = "~sm.ns(Pseudotime, df = 3)", 
                                                 cores = parallel::detectCores() - 2)

# Filter genes based on FDR
pseudo.maturation.diff.filtered <- pseudo.maturation.diff %>% filter(qval < 1e-40)

Direction of the DEG by calculating the area between curves (ABC)

Smooth commun genes along the two trajectories

# Create a new pseudo-DV vector of 200 points
nPoints <- 200

new_data = list()
for (Lineage in unique(pData(gbm_cds)$Lineage)){
  new_data[[length(new_data) + 1]] = data.frame(Pseudotime = seq(min(pData(gbm_cds)$Pseudotime), max(pData(gbm_cds)$Pseudotime), length.out = nPoints), Lineage=Lineage)
}

new_data = do.call(rbind, new_data)

# Smooth gene expression
Diff.curve_matrix <- genSmoothCurves(gbm_cds[as.character(pseudo.maturation.diff.filtered$gene_short_name),],
                                      trend_formula = "~sm.ns(Pseudotime, df = 3)*Lineage",
                                      relative_expr = TRUE,
                                      new_data = new_data,
                                      cores= parallel::detectCores() - 2)

Compute the ABC for each gene

# Extract matrix containing smoothed curves for each lineages
Pal_curve_matrix <- Diff.curve_matrix[, 1:nPoints] #Pallial points
CR_curve_matrix <- Diff.curve_matrix[, (nPoints + 1):(2 * nPoints)] #CR points

# Direction of the comparison : postive ABCs <=> Upregulated in CR lineage
ABCs_res <- CR_curve_matrix - Pal_curve_matrix

# Average logFC between the 2 curves
ILR_res <- log2(CR_curve_matrix/ (Pal_curve_matrix + 0.1)) 
  
ABCs_res <- apply(ABCs_res, 1, function(x, nPoints) {
                  avg_delta_x <- (x[1:(nPoints - 1)] + x[2:(nPoints)])/2
                  step <- (100/(nPoints - 1))
                  res <- round(sum(avg_delta_x * step), 3)
                  return(res)},
                  nPoints = nPoints) # Compute the area below the curve
  
ABCs_res <- cbind(ABCs_res, ILR_res[,ncol(ILR_res)])
colnames(ABCs_res)<- c("ABCs", "Endpoint_ILR")

# Import ABC values into the DE test results table
pseudo.maturation.diff.filtered <- cbind(pseudo.maturation.diff.filtered[,1:4],
                                         ABCs_res,
                                         pseudo.maturation.diff.filtered[,5:6])

Cajal-Retzius cells specific trajectory analysis

# Extract Cajal-Retzius expressed genes
CR.res <- as.data.frame(pseudo.maturation.diff.filtered[pseudo.maturation.diff.filtered$ABCs > 0,])
CR.genes <- row.names(CR.res)

CR_curve_matrix <- CR_curve_matrix[CR.genes, ]

Gene expression profiles along the trajectory

## Cluster gene by expression profiles
Pseudotime.genes.clusters <- cluster::pam(as.dist((1 - cor(Matrix::t(CR_curve_matrix),method = "pearson"))), k= 5)

CR.Gene.dynamique <- data.frame(Gene= names(Pseudotime.genes.clusters$clustering),
                                 Waves= Pseudotime.genes.clusters$clustering,
                                 Gene.Clusters = Pseudotime.genes.clusters$clustering,
                                 q.val = CR.res$qval,
                                 ABCs= CR.res$ABCs
                                 ) %>% arrange(Gene.Clusters)

row.names(CR.Gene.dynamique) <- CR.Gene.dynamique$Gene
CR.Gene.dynamique$Gene.Clusters <- paste0("Clust.", CR.Gene.dynamique$Gene.Clusters)

# Order the rows using seriation
dst <- as.dist((1-cor(scale(t(CR_curve_matrix)), method = "pearson")))
row.ser <- seriation::seriate(dst, method ="R2E") #"R2E" #TSP #"GW" "GW_ward"
gene.order <- rownames(CR_curve_matrix[get_order(row.ser),])

# Set annotation colors
pal <- wes_palette("Darjeeling1")
anno.colors <- list(lineage = c(Pallial_neurons="#026c9a", Cajal_Retzius="#cc391b"),
                    Gene.Clusters = c(Clust.1 =pal[1] , Clust.2=pal[2], Clust.3=pal[3], Clust.4=pal[4], Clust.5=pal[5]))


pheatmap::pheatmap(Diff.curve_matrix[gene.order,
                                c(200:1,#Pal 
                                  201:400)], #CR
                   scale = "row",
                   cluster_rows = F,
                   cluster_cols = F,
                   annotation_row = CR.Gene.dynamique %>% dplyr::select(Gene.Clusters),
                   annotation_col = data.frame(lineage = rep(c("Pallial_neurons","Cajal_Retzius"), each=200)),
                   annotation_colors = anno.colors,
                   show_colnames = F,
                   show_rownames = F,
                   fontsize_row = 8,
                   color =  viridis::viridis(9),
                   breaks = seq(-2.5,2.5, length.out = 9),
                   main = "")

anno.colors <- list(Cell.state = c(Cycling_RG="#046c9a", Differentiating_cells="#ebcb2e"),
                    Gene.Clusters = c(Clust.1 =pal[1] , Clust.2=pal[2], Clust.3=pal[3], Clust.4=pal[4], Clust.5=pal[5]))

col.anno <- data.frame(Cell.state = rep(c("Cycling_RG","Differentiating_cells"), each=100))
rownames(col.anno) <- 201:400

pheatmap::pheatmap(CR_curve_matrix[gene.order,],
                   scale = "row",
                   cluster_rows = F,
                   cluster_cols = F,
                   annotation_row = CR.Gene.dynamique %>% dplyr::select(Gene.Clusters),
                   annotation_col = col.anno,
                   annotation_colors = anno.colors,
                   gaps_col = 100,
                   show_colnames = F,
                   show_rownames = F,
                   fontsize_row = 8,
                   color =  viridis::viridis(9),
                   breaks = seq(-2.5,2.5, length.out = 9),
                   main = "")

### Gene cluster trend

source("../Functions/functions_GeneClusterTrend.R")

Plot.clust.trends(Neuro.trajectories,
                   Lineage = "Hem",
                   Which.cluster = 1:5,
                   clust.list = Pseudotime.genes.clusters$clustering,
                   Smooth.method = "gam")

GO term enrichment in gene clusters using gprofiler2

CR.gostres <- gost(query = list("Clust.1" = CR.Gene.dynamique %>% filter(Gene.Clusters == "Clust.1") %>% pull(Gene) %>% as.character(),
                             "Clust.2" = CR.Gene.dynamique %>% filter(Gene.Clusters == "Clust.2") %>% pull(Gene) %>% as.character(),
                             "Clust.3" = CR.Gene.dynamique %>% filter(Gene.Clusters == "Clust.3") %>% pull(Gene) %>% as.character(),
                             "Clust.4" = CR.Gene.dynamique %>% filter(Gene.Clusters == "Clust.4") %>% pull(Gene) %>% as.character(),
                             "Clust.5" = CR.Gene.dynamique %>% filter(Gene.Clusters == "Clust.5") %>% pull(Gene) %>% as.character()),
                organism = "mmusculus", ordered_query = F, 
                multi_query = F, significant = T, exclude_iea = T, 
                measure_underrepresentation = F, evcodes = T, 
                user_threshold = 0.05, correction_method = "fdr", 
                domain_scope = "annotated", custom_bg = NULL, 
                numeric_ns = "", sources = c("GO:MF", "GO:BP"), as_short_link = F)

DNA_damage_GOterm <- CR.gostres$result[CR.gostres$result$term_id %in% c("GO:0008630", "GO:0030330", "GO:0031571", "GO:0006974", "GO:0006977","GO:0033554",
                                                                                 "GO:0044773", "GO:0042771", "GO:0042770", "GO:2001021", "GO:1902229"),]

DNA_damage_GOterm[,c(1,2,3,5,6,7,11)]

## # A tibble: 8 × 7
##   query   significant p_value query_size intersection_size precision term_name  
##   <chr>   <lgl>         <dbl>      <int>             <int>     <dbl> <chr>      
## 1 Clust.1 TRUE        0.0172          93                 3    0.0323 intrinsic …
## 2 Clust.2 TRUE        0.00622        182                28    0.154  cellular r…
## 3 Clust.4 TRUE        0.0159         121                 4    0.0331 DNA damage…
## 4 Clust.4 TRUE        0.0283         121                 4    0.0331 negative r…
## 5 Clust.4 TRUE        0.0288         121                21    0.174  cellular r…
## 6 Clust.4 TRUE        0.0339         121                 3    0.0248 regulation…
## 7 Clust.4 TRUE        0.0367         121                 5    0.0413 signal tra…
## 8 Clust.4 TRUE        0.0397         121                 2    0.0165 DNA damage…

CR.gostres <- gost(query = as.character(CR.Gene.dynamique$Gene),
                organism = "mmusculus", ordered_query = F, 
                multi_query = F, significant = T, exclude_iea = T, 
                measure_underrepresentation = F, evcodes = T, 
                user_threshold = 0.05, correction_method = "fdr", 
                domain_scope = "annotated", custom_bg = NULL, 
                numeric_ns = "", sources = c("GO:MF", "GO:BP"), as_short_link = F)

DNA_damage_GOterm <- CR.gostres$result[CR.gostres$result$term_id %in% c("GO:0008630", "GO:0030330", "GO:0031571", "GO:0006974", "GO:0006977",
                                                                                 "GO:0044773", "GO:0042771", "GO:0042770", "GO:2001021", "GO:1902229"),]

DNA_damage_GOterm[,c(1,2,3,5,6,7,11)]

## # A tibble: 10 × 7
##    query   significant p_value query_size intersection_size precision term_name 
##    <chr>   <lgl>         <dbl>      <int>             <int>     <dbl> <chr>     
##  1 query_1 TRUE        0.00107        510                11   0.0216  intrinsic…
##  2 query_1 TRUE        0.00430        510                 7   0.0137  DNA damag…
##  3 query_1 TRUE        0.00505        510                 4   0.00784 mitotic G…
##  4 query_1 TRUE        0.0118         510                30   0.0588  cellular …
##  5 query_1 TRUE        0.0118         510                 3   0.00588 DNA damag…
##  6 query_1 TRUE        0.0184         510                 6   0.0118  mitotic D…
##  7 query_1 TRUE        0.0207         510                 5   0.00980 intrinsic…
##  8 query_1 TRUE        0.0365         510                 9   0.0176  signal tr…
##  9 query_1 TRUE        0.0405         510                 6   0.0118  negative …
## 10 query_1 TRUE        0.0496         510                 4   0.00784 regulatio…

Pallial neuron trajectory analysis

# Extract Pallial neurons trajectory genes
Pal.res <- as.data.frame(pseudo.maturation.diff.filtered[pseudo.maturation.diff.filtered$ABCs < 0,])
Pal.genes <- row.names(Pal.res)

Pal_curve_matrix <- Pal_curve_matrix[Pal.genes, ]

Gene expression profiles along the trajectory

## Cluster gene by expression profiles
Pseudotime.genes.clusters <- cluster::pam(as.dist((1 - cor(Matrix::t(Pal_curve_matrix),method = "spearman"))), k= 5)

Pal.Gene.dynamique <- data.frame(Gene= names(Pseudotime.genes.clusters$clustering),
                             Waves= Pseudotime.genes.clusters$clustering,
                             Gene.Clusters = Pseudotime.genes.clusters$clustering,
                             q.val = Pal.res$pval,
                             ABCs= Pal.res$ABCs
                             ) %>% arrange(Gene.Clusters)

row.names(Pal.Gene.dynamique) <- Pal.Gene.dynamique$Gene
Pal.Gene.dynamique$Gene.Clusters <- paste0("Clust.", Pal.Gene.dynamique$Gene.Clusters)

# Order the rows using seriation
dst <- as.dist((1-cor(scale(t(Pal_curve_matrix)), method = "spearman")))
row.ser <- seriation::seriate(dst, method ="R2E") #"R2E" #TSP #"GW" "GW_ward"
gene.order <- rownames(Pal_curve_matrix[get_order(row.ser),])

# Set annotation colors
pal <- wes_palette("Darjeeling1")
anno.colors <- list(lineage = c(Pallial_neurons="#026c9a", Cajal_Retzius="#cc391b"),
                    Gene.Clusters = c(Clust.1 =pal[1] , Clust.2=pal[2], Clust.3=pal[3], Clust.4=pal[4], Clust.5=pal[5]))


pheatmap::pheatmap(Diff.curve_matrix[gene.order,
                                c(200:1,#Pal
                                  201:400)], #CR
                   scale = "row",
                   cluster_rows = F,
                   cluster_cols = F,
                   annotation_row = Pal.Gene.dynamique %>% dplyr::select(Gene.Clusters),
                   annotation_col = data.frame(lineage = rep(c("Pallial_neurons","Cajal_Retzius"), each=200)),
                   annotation_colors = anno.colors,
                   show_colnames = F,
                   show_rownames = F,
                   fontsize_row = 8,
                   color =  viridis::viridis(9),
                   breaks = seq(-2.5,2.5, length.out = 9),
                   main = "")

anno.colors <- list(Cell.state = c(Cycling_RG="#046c9a", Differentiating_cells="#ebcb2e"),
                    Gene.Clusters = c(Clust.1 =pal[1] , Clust.2=pal[2], Clust.3=pal[3], Clust.4=pal[4], Clust.5=pal[5]))

col.anno <- data.frame(Cell.state = rep(c("Differentiating_cells","Cycling_RG"), each=100))
rownames(col.anno) <- 200:1

pheatmap::pheatmap(Pal_curve_matrix[gene.order,200:1],
                   scale = "row",
                   cluster_rows = F,
                   cluster_cols = F,
                   annotation_row = Pal.Gene.dynamique %>% dplyr::select(Gene.Clusters),
                   annotation_col = col.anno,
                   annotation_colors = anno.colors,
                   gaps_col = 100,
                   show_colnames = F,
                   show_rownames = F,
                   fontsize_row = 8,
                   color =  viridis::viridis(9),
                   breaks = seq(-2.5,2.5, length.out = 9),
                   main = "")

### Gene cluster trend

Plot.clust.trends(Neuro.trajectories,
                   Lineage = "Pallial",
                   Which.cluster = 1:5,
                   clust.list = Pseudotime.genes.clusters$clustering,
                   Smooth.method = "gam")

Session Info

#date
format(Sys.time(), "%d %B, %Y, %H,%M")

## [1] "07 décembre, 2021, 18,44"

#Packages used
sessionInfo()

## R version 4.1.2 (2021-11-01)
## Platform: x86_64-pc-linux-gnu (64-bit)
## Running under: Ubuntu 20.04.3 LTS
## 
## Matrix products: default
## BLAS:   /usr/lib/x86_64-linux-gnu/blas/libblas.so.3.9.0
## LAPACK: /usr/lib/x86_64-linux-gnu/lapack/liblapack.so.3.9.0
## 
## locale:
##  [1] LC_CTYPE=fr_FR.UTF-8       LC_NUMERIC=C              
##  [3] LC_TIME=fr_FR.UTF-8        LC_COLLATE=fr_FR.UTF-8    
##  [5] LC_MONETARY=fr_FR.UTF-8    LC_MESSAGES=fr_FR.UTF-8   
##  [7] LC_PAPER=fr_FR.UTF-8       LC_NAME=C                 
##  [9] LC_ADDRESS=C               LC_TELEPHONE=C            
## [11] LC_MEASUREMENT=fr_FR.UTF-8 LC_IDENTIFICATION=C       
## 
## attached base packages:
## [1] splines   stats4    stats     graphics  grDevices utils     datasets 
## [8] methods   base     
## 
## other attached packages:
##  [1] wesanderson_0.3.6   cowplot_1.1.1       ggExtra_0.9        
##  [4] RColorBrewer_1.1-2  dplyr_1.0.7         seriation_1.3.1    
##  [7] gprofiler2_0.2.1    monocle_2.22.0      DDRTree_0.1.5      
## [10] irlba_2.3.3         VGAM_1.1-5          ggplot2_3.3.5      
## [13] Biobase_2.54.0      BiocGenerics_0.40.0 Matrix_1.3-4       
## [16] princurve_2.1.6     SeuratObject_4.0.4  Seurat_4.0.5       
## 
## loaded via a namespace (and not attached):
##   [1] plyr_1.8.6            igraph_1.2.9          lazyeval_0.2.2       
##   [4] densityClust_0.3      listenv_0.8.0         scattermore_0.7      
##   [7] fastICA_1.2-3         digest_0.6.29         foreach_1.5.1        
##  [10] htmltools_0.5.2       viridis_0.6.2         fansi_0.5.0          
##  [13] magrittr_2.0.1        tensor_1.5            cluster_2.1.2        
##  [16] ROCR_1.0-11           limma_3.50.0          globals_0.14.0       
##  [19] matrixStats_0.61.0    docopt_0.7.1          spatstat.sparse_2.0-0
##  [22] colorspace_2.0-2      ggrepel_0.9.1         xfun_0.28            
##  [25] RCurl_1.98-1.5        sparsesvd_0.2         crayon_1.4.2         
##  [28] jsonlite_1.7.2        spatstat.data_2.1-0   survival_3.2-13      
##  [31] zoo_1.8-9             iterators_1.0.13      glue_1.5.1           
##  [34] polyclip_1.10-0       registry_0.5-1        gtable_0.3.0         
##  [37] leiden_0.3.9          future.apply_1.8.1    abind_1.4-5          
##  [40] scales_1.1.1          pheatmap_1.0.12       DBI_1.1.1            
##  [43] miniUI_0.1.1.1        Rcpp_1.0.7            viridisLite_0.4.0    
##  [46] xtable_1.8-4          reticulate_1.22       spatstat.core_2.3-1  
##  [49] htmlwidgets_1.5.4     httr_1.4.2            FNN_1.1.3            
##  [52] ellipsis_0.3.2        ica_1.0-2             farver_2.1.0         
##  [55] pkgconfig_2.0.3       sass_0.4.0            uwot_0.1.10          
##  [58] deldir_1.0-6          utf8_1.2.2            labeling_0.4.2       
##  [61] tidyselect_1.1.1      rlang_0.4.12          reshape2_1.4.4       
##  [64] later_1.3.0           munsell_0.5.0         tools_4.1.2          
##  [67] cli_3.1.0             generics_0.1.1        ggridges_0.5.3       
##  [70] evaluate_0.14         stringr_1.4.0         fastmap_1.1.0        
##  [73] yaml_2.2.1            goftest_1.2-3         knitr_1.36           
##  [76] fitdistrplus_1.1-6    purrr_0.3.4           RANN_2.6.1           
##  [79] pbapply_1.5-0         future_1.23.0         nlme_3.1-153         
##  [82] mime_0.12             slam_0.1-49           rstudioapi_0.13      
##  [85] compiler_4.1.2        plotly_4.10.0         png_0.1-7            
##  [88] spatstat.utils_2.2-0  tibble_3.1.6          bslib_0.3.1          
##  [91] stringi_1.7.6         highr_0.9             lattice_0.20-45      
##  [94] HSMMSingleCell_1.14.0 vctrs_0.3.8           pillar_1.6.4         
##  [97] lifecycle_1.0.1       spatstat.geom_2.3-0   combinat_0.0-8       
## [100] lmtest_0.9-39         jquerylib_0.1.4       RcppAnnoy_0.0.19     
## [103] bitops_1.0-7          data.table_1.14.2     httpuv_1.6.3         
## [106] patchwork_1.1.1       R6_2.5.1              promises_1.2.0.1     
## [109] TSP_1.1-11            KernSmooth_2.23-20    gridExtra_2.3        
## [112] parallelly_1.29.0     codetools_0.2-18      MASS_7.3-54          
## [115] assertthat_0.2.1      withr_2.4.3           qlcMatrix_0.9.7      
## [118] sctransform_0.3.2     mgcv_1.8-38           parallel_4.1.2       
## [121] grid_4.1.2            rpart_4.1-15          tidyr_1.1.4          
## [124] rmarkdown_2.11        Rtsne_0.15            shiny_1.7.1

---

1. Institute of Psychiatry and Neuroscience of Paris, INSERM U1266, 75014, Paris, France, matthieu.moreau@inserm.fr↩︎